Item 12 · synthetic

Synthetic Strong Fit Security Endpoint

scenario synthetic_strong_fit_security_endpoint

Input

Marco Velasquez
Director of Engineering at Sentinova

I've spent the last three years at Sentinova helping build out the engineering org from a scrappy 15-person team to close to 180 people spread across San Francisco and Warsaw. We closed our Series C in early 2023, and the growth since then has been real — new hires, new architecture decisions, and a whole lot of hard conversations about how to scale without losing the quality bar we set early on.

My focus is on the platform and detection teams. We ship an endpoint security agent that streams telemetry — process trees, network events, file system changes — into a classification pipeline that uses a fine-tuned transformer model to identify threat patterns in near real-time. Getting that inference latency down to something operationally useful was a project I led for most of last year, working closely with our ML and infrastructure engineers.

Before Sentinova I was a senior engineering manager at Cylance for four years, and before that a software engineer at Palo Alto Networks. I'm comfortable at the intersection of security domain knowledge and distributed systems, which is where most of the interesting problems in this space live.

I'm based in the Bay Area and occasionally write about ML deployment tradeoffs on my personal site. Happy to connect with engineers and founders working on detection, response, or applied ML in security contexts.

Sentinova is a B2B SaaS endpoint security company headquartered in San Francisco, with an engineering hub in Warsaw, Poland. Founded in 2018, the company raised a Series C round in early 2023 and employs approximately 175 people. Sentinova's platform collects endpoint telemetry — process, network, and file-system events — and runs it through a proprietary AI threat-classification pipeline to detect malicious activity with low latency. Customers are mid-market and enterprise security operations teams in North America and Europe. Annual recurring revenue is estimated in the mid-to-upper tens of millions of dollars range. The company sells directly and through a small number of MSSP partners.

Gold

Expected action: auto_add
Fit score: 0.92
Industry: B2B SaaS
Segment: Security / endpoint protection
Seniority: Director
Company size: 51-200
Language: en
Notes: Strong fit. ARR sits near the $50M ceiling but the labeller's best inference is inside the range, matching exemplar 4's posture.

Predictions

Integratedpass

Action: auto_add ✓
Fit: 0.82 (gold 0.92, |Δ| 0.10)
Industry: B2B SaaS ✓
Segment: Endpoint Security / Cybersecurity ✗
Seniority: Director ✓
Company size: 51-200 ✓
Grounding: 7/7 claims grounded (100.0%)

Claims (7)

Marco Velasquez holds the Director of Engineering role at Sentinova, an ICP-target seniority level.
✓ in input Director of Engineering at Sentinova
Sentinova is a B2B SaaS company, matching the ICP product shape requirement.
✓ in input Sentinova is a B2B SaaS endpoint security company
Sentinova closed a Series C round in early 2023, placing it squarely within the ICP funding stage range.
✓ in input We closed our Series C in early 2023
Sentinova's headcount of approximately 175–180 people sits comfortably within the ICP's 20–250 range.
✓ in input close to 180 people spread across San Francisco and Warsaw
Sentinova ships a live AI feature — a fine-tuned transformer model for real-time threat classification — satisfying the ICP's product-shape criterion.
✓ in input a classification pipeline that uses a fine-tuned transformer model to identify threat patterns in near real-time
ARR is estimated in the mid-to-upper tens of millions, which may approach or exceed the ICP's $50M ceiling, introducing some scoring uncertainty.
✓ in input Annual recurring revenue is estimated in the mid-to-upper tens of millions of dollars range
Marco personally led the project to reduce inference latency on the transformer-based detection pipeline, demonstrating hands-on ownership of AI deployment decisions.
✓ in input Getting that inference latency down to something operationally useful was a project I led for most of last year, working closely with our ML and infrastructure engineers

Draft hook

Hi Marco — saw that you spent much of last year driving down inference latency on Sentinova's fine-tuned transformer threat-classification pipeline. That's exactly the kind of production AI deployment challenge we work on with Director- and VP-level engineering leaders at Series B/C security companies. Would love to share one specific pattern that's helped teams in your position shave latency without retraining. Worth a 20-minute call?

Reasoning

All five ICP dimensions are evaluable from the input. Stage (Series C), headcount (~175–180), product shape (live fine-tuned transformer in production), and role (Director of Engineering) are all strong matches scoring 0.90–1.0. The only soft spot is ARR: the company text says "mid-to-upper tens of millions," which could sit at or above the $50M ICP ceiling, yielding a 0.55 ARR dimension score. However, the four remaining dimensions are so well-aligned — including an exact role match and an unusually clear AI-feature signal — that the holistic score reaches 0.82. Every claim in the payload is anchored to a verbatim quote from the input. Because the score exceeds 0.80 and all claims are grounded, the action is auto_add.

Chatmiss

1/8 claim source quote(s) not in input

Action: auto_add ✓
Fit: 0.85 (gold 0.92, |Δ| 0.07)
Industry: B2B SaaS ✓
Segment: Endpoint Security / AI-powered threat detection ✗
Seniority: Director ✓
Company size: 51-200 ✓
Grounding: 7/8 claims grounded (87.5%)

Claims (8)

Marco is Director of Engineering at Sentinova responsible for platform and detection teams
✓ in input My focus is on the platform and detection teams.
He led the inference latency optimization work on Sentinova's transformer-based threat classification pipeline
✓ in input Getting that inference latency down to something operationally useful was a project I led for most of last year, working closely with our ML and infrastructure engineers.
Sentinova closed Series C in early 2023
✓ in input We closed our Series C in early 2023
Sentinova grew engineering from 15 to approximately 180 people over three years
✓ in input I've spent the last three years at Sentinova helping build out the engineering org from a scrappy 15-person team to close to 180 people
Marco has previous experience as Senior Engineering Manager at Cylance and Software Engineer at Palo Alto Networks
✓ in input Before Sentinova I was a senior engineering manager at Cylance for four years, and before that a software engineer at Palo Alto Networks.
He publicly writes about ML deployment tradeoffs in security
✗ not in input occasionally writes about ML deployment tradeoffs on my personal site
Sentinova's platform streams endpoint telemetry through an AI classification pipeline for threat detection
✓ in input We ship an endpoint security agent that streams telemetry — process trees, network events, file system changes — into a classification pipeline that uses a fine-tuned transformer model
Engineering team is distributed across San Francisco and Warsaw
✓ in input spread across San Francisco and Warsaw

Draft hook

Marco — saw you've been writing about ML deployment tradeoffs. The work you led getting threat-classification latency down to operationally useful thresholds is exactly the kind of problem we focus on. One thing that tends to get harder as the eng org scales — which yours clearly has — is maintaining visibility into why latency regresses in production. Model updates, infra changes, and data drift all look the same in an aggregate p99 metric. Worth a quick conversation to see if what we're building is relevant?

Reasoning

Marco is an exceptionally strong ICP fit across all five dimensions. He's a Director of Engineering (target role) at a Series C B2B SaaS company (~175 headcount, which is squarely in the 51-200 range). Sentinova's product is an AI-powered threat detection pipeline — the core offering is the ML system, not ancillary to it. The only mild flag is ARR positioning at the upper edge of the ICP range, but the notes indicate "probably at the upper edge" is acceptable given the strength of other dimensions. More importantly, Marco has direct hands-on ownership of the exact problem (inference latency optimization) that would likely resonate with a sales pitch. His security background (PANw → Cylance → Sentinova) means he's a technically credible buyer or influencer on security tooling. His public writing signals engagement and thought leadership. The distributed SF/Warsaw team structure adds operational complexity that often surfaces hidden pain points. This is not a title match alone — this is a lead with demonstrable decision-making authority over the specific technical domain (ML inference, platform architecture, threat detection) and personal experience with the scaling challenges that typically drive buying decisions in this space. Auto-add is justified."